Early-Token Confidence Predicts Reasoning Quality in Multi-Agent LLM Debate

• URL: http://arxiv.org/abs/2606.10307v1
• Date: Tue, 09 Jun 2026 01:52:59 GMT
• Title: Early-Token Confidence Predicts Reasoning Quality in Multi-Agent LLM Debate
• Abstract summary: We investigate whether intrinsic confidence signals, token-level log-probabilities from decoding, can predict reasoning quality.<n>We find that early-token confidence, particularly within the first few generated tokens, is consistently the strongest predictor of reasoning quality.
• Score: 2.1987782243719898
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Evaluating reasoning quality in multi-agent LLM systems is challenging, especially for open-ended tasks without reference answers. We investigate whether intrinsic confidence signals, token-level log-probabilities from decoding, can predict reasoning quality as assessed by LLM-as-judge evaluation. Using a debate-based essay scoring framework, we compare confidence proxies against rubric-based judge scores across two ASAP essay sets. We find that early-token confidence, particularly within the first few generated tokens, is consistently the strongest predictor of reasoning quality, outperforming full-sequence statistics. Analysis of log-probability trajectories shows that the opening phase of generation is the most heterogeneous and therefore most informative. We also observe a systematic asymmetry between agent roles, with stronger alignment between confidence and quality for supportive reasoning than for adversarial critique. These results suggest that early decoding dynamics provide a lightweight and effective signal for estimating reasoning reliability in multi-agent LLM systems.

Related papers

• The Confident Liar: Diagnosing Multi-Agent Debate with Log-Probabilities and LLM-as-Judge [2.1987782243719898]
  This paper studies the relationship between three signals in multi-agent debate.<n>We examine whether internal confidence signals predict externally evaluated reasoning quality.<n>We find that confidence aligns with judged reasoning quality roughly twice as strongly for the Constructor as for the Auditor.
  arXiv  Detail & Related papers  (2026-06-09T01:33:15Z)
• Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents? [61.49434544687523]
  We introduce REFLECT, a meta-evaluation benchmark targeting fine-grained failure detection in agentic environments.<n> REFLECT defines a detailed taxonomy of process- and outcome-level failure modes, instantiated by performing controlled and localized interventions.<n>Our experiments show that even the best-performing models achieve overall accuracies below 55% across reasoning, tool-use, and report-quality failures.
  arXiv  Detail & Related papers  (2026-05-18T23:55:08Z)
• Robust Checkpoint Selection for Multimodal LLMs via Agentic Evaluation and Stability-Aware Ranking [0.4562811934651716]
  We formulate checkpoint selection as a robust decision problem under evaluation uncertainty.<n>We propose a multi-stage framework that integrates curated real-world data, structured LLM-based judgment, and multi-stage ranking protocols.<n>We demonstrate that data quality, specifically OCR readability, is a critical determinant of evaluation validity.
  arXiv  Detail & Related papers  (2026-05-13T12:18:32Z)
• Epistemic Context Learning: Building Trust the Right Way in LLM-Based Multi-Agent Systems [94.9141394384021]
  Individual agents in multi-agent systems often lack robustness, tending to blindly conform to misleading peers.<n>We show this weakness stems from both sycophancy and inadequate ability to evaluate peer reliability.<n>We first formalize the learning problem of history-aware reference, introducing the historical interactions of peers as additional input.<n>We then develop Epistemic Context Learning (ECL), a reasoning framework that conditions predictions on explicitly-built peer profiles from history.
  arXiv  Detail & Related papers  (2026-01-29T13:59:32Z)
• Gaming the Judge: Unfaithful Chain-of-Thought Can Undermine Agent Evaluation [76.5533899503582]
  Large language models (LLMs) are increasingly used as judges to evaluate agent performance.<n>We show this paradigm implicitly assumes that the agent's chain-of-thought (CoT) reasoning faithfully reflects both its internal reasoning and the underlying environment state.<n>We demonstrate that manipulated reasoning alone can inflate false positive rates of state-of-the-art VLM judges by up to 90% across 800 trajectories spanning diverse web tasks.
  arXiv  Detail & Related papers  (2026-01-21T06:07:43Z)
• Rubric-Conditioned LLM Grading: Alignment, Uncertainty, and Robustness [4.129847064263056]
  We systematically evaluate the performance of Large Language Models for rubric-based short-answer grading.<n>We find that alignment is strong for binary tasks but degrades with increased rubric granularity.<n>Experiments reveal that while the model is resilient to prompt injection, it is sensitive to synonym substitutions.
  arXiv  Detail & Related papers  (2025-12-21T05:22:04Z)
• Self-Evaluating LLMs for Multi-Step Tasks: Stepwise Confidence Estimation for Failure Detection [1.1087735229999818]
  Self-evaluating large language models (LLMs) provides meaningful confidence estimates in complex reasoning.<n>Stepwise evaluation generally outperforms holistic scoring in detecting potential errors.
  arXiv  Detail & Related papers  (2025-11-10T18:19:51Z)
• FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning [62.452350134196934]
  FaithCoT-Bench is a unified benchmark for instance-level CoT unfaithfulness detection.<n>Our framework formulates unfaithfulness detection as a discriminative decision problem.<n>FaithCoT-Bench sets a solid basis for future research toward more interpretable and trustworthy reasoning in LLMs.
  arXiv  Detail & Related papers  (2025-10-05T05:16:54Z)
• LLM-as-a-Judge: Rapid Evaluation of Legal Document Recommendation for Retrieval-Augmented Generation [40.06592175227558]
  This paper investigates a principled approach to evaluating Retrieval-Augmented Generation systems in legal contexts.<n>We find that traditional agreement metrics like Krippendorff's alpha can be misleading in the skewed distributions typical of AI system evaluations.<n>Our findings suggest a path toward scalable, cost-effective evaluation that maintains the precision demanded by legal applications.
  arXiv  Detail & Related papers  (2025-09-15T19:20:21Z)
• Attention Knows Whom to Trust: Attention-based Trust Management for LLM Multi-Agent Systems [52.57826440085856]
  Large Language Model-based Multi-Agent Systems (LLM-MAS) have demonstrated strong capabilities in solving complex tasks but remain vulnerable when agents receive unreliable messages.<n>This vulnerability stems from a fundamental gap: LLM agents treat all incoming messages equally without evaluating their trustworthiness.<n>We propose Attention Trust Score (A-Trust), a lightweight, attention-based method for evaluating message trustworthiness.
  arXiv  Detail & Related papers  (2025-06-03T07:32:57Z)
• PredictaBoard: Benchmarking LLM Score Predictability [50.47497036981544]
  Large Language Models (LLMs) often fail unpredictably.<n>This poses a significant challenge to ensuring their safe deployment.<n>We present PredictaBoard, a novel collaborative benchmarking framework.
  arXiv  Detail & Related papers  (2025-02-20T10:52:38Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.